Summary
Daniar Kurniawan is a Machine Learning Systems Engineer with 12 years of experience at the intersection of ML and distributed systems, currently optimizing ML deployments using DPU, GPU-over-RDMA and NVMe technologies at MangoBoost. He holds a PhD from the University of Chicago and during his graduate work built novel caching and I/O admission algorithms that cut memory use and p99 tail latencies dramatically, and developed inference methods with up to 10x latency improvements. His research track includes publications and systems work on detecting distributed concurrency bugs, resource-allocation improvements for cloud services, and practical integrations of ML into storage systems like Ceph. Daniar has deep hands-on experience hacking large-scale systems—Cassandra, Apache Ignite, Voldemort, Spark—and has repeatedly shipped firmware and cluster-scale optimizations at Seagate and Microsoft collaborations. Notably, he implemented a PyTorch-integrated caching approach that reduced Meta-style deep learning memory footprints by up to 94%, and prototyped kernel-level NN deployments achieving microsecond inference latencies. Based in Bellevue, WA, he blends rigorous academic research with production-grade systems engineering to push low-latency, high-throughput ML serving.
12 years of coding experience
9 years of employment as a software developer
Bachelor of Engineering (B.Eng.) Computer Science, Bachelor of Engineering (B.Eng.) Computer Science at Institut Teknologi Bandung
Doctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at University of Chicago
Indonesian, English